Proper way to ipsec filtering

From: "Kamil Jońca" <kjonca@op.pl>
To: netfilter@vger.kernel.org
Subject: Proper way to ipsec filtering
Date: Sun, 17 Apr 2022 08:37:47 +0200	[thread overview]
Message-ID: <878rs48bb8.fsf@alfa.kjonca> (raw)

What is the best way to create rules used to ipsec traffic filtering?

So far I have bunch rules created per reqid like that:
table ip filter { # handle 13

chain INPUT { # handle 1
                type filter hook input priority filter; policy drop;
                iif "eth0" ipsec in reqid 1 counter packets 100672 bytes 11492891 jump ipsec-in-1 comment "ed19af3c-f504-11e9-b59d-00e081736ba6/1/in" # handle 326
                [...]

 }

[...]
        chain ipsec-in-1 { # handle 323
                ip saddr yyy ip daddr xxxx/24 counter packets 50871 bytes 5614784 jump c1 # handle 325
                ip protocol ipencap ip daddr zzzz counter packets 49801 bytes 5878107 accept # handle 324
        }

}

And insert / remove rules from INPUT  (and add / delete ipsec-in-*
chains) during connecting disconnecting clients.
This was I configured when I migrated from iptables some time ago.
But I believe this is not the best method for nftables. So has anybody
suggestion what is the best practicte to handle this situation?
I tried to use maps/vmaps but reqid cannot be use as index.

Am I missing something?
KJ

-- 
http://stopstopnop.pl/stop_stopnop.pl_o_nas.html